Saliency Detection Via Ranking With Reconstruction Error

Nowadays with the rapid development of science and technology, smart phone and internet bring great facilities to people, for more and more pictures emerging. A rapid, precise and effective method is badly in need of face up to the mass images. More attention is paid to saliency detection because of its widespread use. Recent years saliency detection plays an important role in computer vision.In the paper, we present a bottom-up salient object detection method. First, we compute two complementary coarse saliency maps based on the dense and sparse reconstruction errors considering the boundary priors. Then we rank the coarse maps and integrate the ranking values in a multiplication way. In the end, we optimize the picture with Gauss blur and ranking twice. Compute an exact saliency map which is robust to the images with cluttered background. Exten-sive experiments on two public available databases demonstrate the superiority of the proposed method in comparison to 12 state-the-of-art approaches.